I have a dataset and website which is partially derived from Wikipedia text. This seems to be licenced CC BY-SA. (This is merged with other free data: OGL, ODbL, etc).

My dataset is CC BY-SA compatible, and I certainly intend to share it so licensed. Everything downstream of the wikipedia data, dataset-wise will be shared. The sui generis database terms are also fine.

However, I'd like to use some stock photos around the site to jolly it up. I have these under a liberal license which permits sharing, but it's an esoteric, custom license (with exciting, extra terms which I'm happy to comply with, seemingly around me not making tea-towels out of their images). I cannot relicense these images CC BY-SA.

The photos are not really integrated with the data at all: they are illustrative, there to jolly things up. As far as I can tell, Wikipedia itself goes much more than this, and distributes their text alongside illustrations with many licenses which directly add value to the text.

I read the discussions on OpenStreetMap's decision to switch to ODbL, and now facing a similar headache, I can see where they are coming from.

So they clearly think this is fine for them. But when I read the legal code, the definition of "Adapted Material" is broad and vague. It seems to me it would cover such a use case.

Later I might want to do other things tangential to the data. For example, I might want to use a non-free font on the site, or elsewhere on the site host a PDF which is SA incompatible.

What is a good rule of thumb so I can know where the red line is?

  • 1
    I withdraw my point about Wikipedia's use of CC images in articles, thanks to @Giacomo Catenazzi 's answer. However my main point stands. How does one draw the line between what the ODbL would call a collection (a "Collective Database" in their context) and an adaptation (a "Derivative Database" in their context) in the context of CC-BY-SA licenced data? This lack of clarity is making me shy of using SA licenced datasets at all, even though it would be less work to abandon the less fundamental SA content rather than find CC-BY/CC0/OGL sources.
    – Dan
    Commented Jun 28 at 20:33

2 Answers 2


We don't have a lot of jurisprudence about the CC licences, but one of the few cases we do have is Drauglis v. Kappa Map Group. In Drauglis, a basically-unmodified CC BY-SA image was used alongside other images on the cover of a road atlas. The court held that neither the atlas as a whole, nor the cover as a whole, constituted derivative works within the meaning of the licence, and thus the Share-Alike obligations extended no further than the cropped copy of the CC BY-SA photo that appeared thereon.

Note that is not a binding judgement anywhere, but as courts tend to follow each other unless they see good reasons not to, it gives us some idea of how other courts might approach the question.

None of the arguments in Drauglis revolved around the fact it was a cover image, nor that it was in print, so using that as a guide, I think you'd be fine using the images as you'd have described.

I'm less sure I'd mix them all into one database, though; we have several questions on here about mixing content under different licences, and it seems generally agreed that while it's perfectly possible, it's extremely helpful to keep differently-licensed content as well-separated as possible. A single database containing content under a number of different licences won't be much help to anyone.

And, as usual, IANAL/IANYL, so take professional advice before you bet your company, or your house, on this analysis.

  • I don't see that putting the images into the database will make a difference. The problem will come if you don't have a correct citation, or you mix thinks up: a derivative set from more than one incompatible licence. Commented Jul 1 at 18:03
  • Looking in parallel elsewhere re database co-mingling, I now agree with you on that. ODbL seem to be CC-BY-SA incompatible on a technicality, so I need to abandon one or the other. Goodness knows what the community is trying to achieve tactically with this confusion.
    – Dan
    Commented Jul 2 at 8:11

Every piece of code has own license (set by copyright holder) and you should apply the term of license. It is something fundamental, but overlooked when we check the derivative work. But look both independently. Derivative work license doesn't erase the license of single components. It is a complement.

So starting from bottom:

It depends on the commercial font: you may get a license that font designer should bet one dollar per every copy printed or distributed. So you see it is SA incompatible: people cannot share the PDF. Usually font licenses are much better, and allows copies (but maybe not extending text). So you may have passed the red-line.

Also we distinguish fonts with lettering. The first is independent from text, so any text is not derivative work. Lettering (e.g. in signs and posters) text and design is linked together, so in such case text may be derivative work from font (think movie posters, translated later in a new language and maybe adding characters).

See: we need to check single license and the license of the derivative work.

For images you will apply same rule: you cannot change license just because you use it in a database or in a text: image author/copyright holder has explicit set a license. You should keep track of each file license. Then you should check if you can include in your SA site, and so you check again all the licenses about copying and derivative work. But such derivative license doesn't change the license of each single file.

As you see, it is something problematic on large scale, so Wikipedia has a minimum requirements, and some institutions/companies are requiring some machine readable license, so to do the check automatically. And strange license gives headaches (so many project try to avoid them and remove content which uses them): wording is important, considering jurisdictions around the world.

Personally: I would check Wikipedia meta on the main used licenses and whitelist the licenses which I can agree and I can use on my derivative work (and probably I would restrict mainly just on CC, public domain). And ignore the rest. I do not think you can generalize, but if all your site is similar (in license) as Wikipedia.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.